Optimizations of a fast multipole symmetric Galerkin boundary element method code

Abstract

This paper presents some optimizations of a fast multipole symmetric Galerkin boundary element method code. Except general optimizations, the code is specially sped up for crack propagation problems. Existing useful computational results are saved and re-used during the propagation. Some time-consuming phases of the code are accelerated by a shared memory parallelization. A new sparse matrix method is designed based on coordinate format and compressed sparse row format to limit the memory required during the matrix construction phase. The remarkable performance of the new code is shown through many simulations including large-scale problems.